Automates Microsoft Excel using AI assistants like GitHub Copilot and Claude via a Model Context Protocol (MCP) server. It enables natural language commands to control a comprehensive suite of Excel functionalities, including Power Query, DAX, VBA, PivotTables, Charts, and formatting, by directly interacting with the Excel COM API. This empowers data analysts, developers, and business users to automate complex workflows and build sophisticated Excel-based solutions efficiently.

How It Works

The project utilizes Microsoft Excel's official COM API to drive the actual Excel application, ensuring live operations such as Power Query refreshes, PivotTable recalculations, and DAX evaluations are performed within the live Excel environment. It offers two primary, first-class interfaces: an MCP Server for conversational AI and a CLI (excelcli) for coding agents and scripting. Both interfaces share a unified core codebase, guaranteeing identical behavior for all 232 operations. The CLI uses a background daemon for persistent sessions, while the MCP Server runs in-process for low-latency automation.

Quick Start & Requirements

• Primary Install/Run: VS Code Extension (recommended), download executables from GitHub releases, or install via .NET tool (dotnet tool install --global Sbroenne.ExcelMcp.McpServer).
• Prerequisites: Windows OS, Microsoft Excel 2016 or later, Desktop Environment (not server-side). Node.js is required for npx commands.
• Links:
  • GitHub Releases: https://github.com/sbroenne/mcp-server-excel/releases/latest
  • MCP Server Installation Guide: (Refer to release assets)
  • CLI Installation Guide: (Refer to release assets)
  • VS Code Extension: Search "ExcelMcp" in the VS Code Marketplace.

Highlighted Details

• Provides 26 specialized tools with 232 operations covering Power Query, Data Model/DAX, Tables, PivotTables, Charts, VBA, Ranges, Worksheets, Connections, Files, and more.
• Directly drives the real Excel engine via COM API, preserving existing workbooks, formulas, and formatting.
• Offers two equal entry points (MCP Server and CLI) with identical operation behavior and a unified codebase.
• The CLI interface is optimized for LLMs, using 64% fewer tokens than the MCP Server.
• Features an "Agent Mode" allowing users to watch AI-driven Excel automation in real-time.

Maintenance & Community

Developed using GitHub Copilot AI assistance. Mentions the Model Context Protocol community and Open Source Community. Related projects include sister MCP servers for PowerPoint, Windows, and OBS Studio. No direct links to Discord/Slack or a public roadmap were found in the README.

Licensing & Compatibility

• License: MIT License.
• Compatibility: Strictly Windows-only due to COM interop. Requires Microsoft Excel 2016 or later. Not suitable for server-side processing or Linux/macOS environments.

Limitations & Caveats

Requires a local installation of Microsoft Excel and runs exclusively on Windows. It is not designed for server-side data processing or high-volume batch operations. Users must close all Excel files before using the server, as it requires exclusive access to workbooks during automation.